Threat Classification on Real-world temporal split 83k samples (test)
68.1AccuracySALM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SALMMethod=Semantic retrieval via contrastive pretraining2026.03 | 68.1 | 30.1 | |
| TF-IDF + RFFeatures=sparse TF-IDF vectors, Classifier=Random Forest2026.03 | 65.7 | 29.8 | |
| FT CodeBERT + MLPBackbone=CodeBERT, Classifier=Multi-layer perceptron (MLP), Training Mode=Fine-tuned2026.03 | 62.3 | 26 | |
| Embedding SimilarityMethod=RAG-style / k-nearest-neighbor majority voting, Backbone=Sentence-BERT (all-MiniLM-L6-v2), Search Index=HNSW2026.03 | 48.9 | 20.9 |